Just came across a recent piece by Carletto et al. in the Journal of Development Economics:

Fact or artifact: The impact of measurement errors on the farm size–productivity relationship

Carletto et al. say:

This paper revisits the role of land measurement error in the inverse farm size and productivity relationship (IR). By making use of data from a nationally representative household survey from Uganda, in which self-reported land size information is complemented by plot measurements collected using Global Position System (GPS) devices we reject the hypothesis that IR may just be a statistical artifact linked to problems with land measurement error. In particular, we explore: (i) what are the determinants of the bias in land measurement, (ii) how this bias varies systematically with plot size and landholding, and (iii) the extent to which land measurement error affects the relative advantage of smallholders implied by the IR. Our findings indicate that using an improved measure of land size strengthens the evidence in support of the existence of the IR.

(Emphasis added.) The Inverse Relationship (IR) argument has been around (at least) since work by Nobel winner Amartya Sen in the 1960s. It has weathered some slings and arrows of meaningful academic critique though I have to say it always strikes me as somewhat curious that it is more often treated as something to be explained away rather than something to be explained. That is, if you read the linked piece by Duke economist Marc Bellamare (a graduate of Cornell’s ag economics program, a program I had peripheral contact with during my time at Cornell), I think one would find it fair to say that he grants that it could be a real and consistent dynamic, but that he treats the possibility as being doubtful. The thing that I find a bit grating is that the usual hypotheses advanced to explain the IR do not include the one I (perhaps somewhat imperiously) view as the most likely: that ecological heterogeneity causes diseconomies of scale. That is, the economies of scale of larger farmers tend to come from homogenization–monocultural fields that can be machine-tended because of regular spacing, genetic uniformity, etc. But if environments–not just soil, mind you, but both biotic and abiotic elements–are typically heterogeneous (which they are) then one might expect that the efforts needed to create and maintain the scale-able homogeneity might detract from total productivity and efficiency. The over-simple example I usually give is that you might have a piece of land consisting of a marshy area, a forest, and a field. In terms of ecological/agricultural productivity, it seems like you might produce more total usable biomass by using the marsh for nutrients (“organic-rich silt“), pest predator source populations (say), fishing, and more; the field for a polyculture of some kind; and the forest for, say, a small population of pigs, small amount of harvestable timber, and moisture/water regulation (as in the infamous Polyface Farm, say). You cannot, of course, easily sell a small amount of fish, timber, marsh muck, squash, beans, and corn to Whole Foods or WinCo. You cannot drive over them all with a tractor, even one with GPS. But you might produce more food per unit area. That is: Joel Salatin‘s method is probably not scaleable in the way we think of most large farms–even though his farm is pretty large, to be fair, it’s also hella-diverse. 

Empirically, I haven’t seen much done with this hypothesis of “mine”. (Really, I draw it from numerous conversations with John Vandermeer while I was doing my PhD.) But it seems highly ecologically plausible, and even quite possibly likely. But even if it’s to be eventually proven wrong, I would like to see it included in the knee-jerk “But have you considered?“